OAI/GroupTheory/PolycyclicRecognition/Analysis/FiniteAffineSelection.lean
5 declarations, 0 contain sorry, 0 declared as axioms
Canonical finite labels and separation from translated competitors
- def
OAI.PolycyclicRecognition.minimizingFiniteLabelline 15Source documentation
Fix one minimizing label at every parameter value.
def minimizingFiniteLabel {L : Type*} [Fintype L] [Nonempty L] (cost : L → ℝ) : L := … - theorem
OAI.PolycyclicRecognition.minimizingFiniteLabel_leline 18 - theorem
OAI.PolycyclicRecognition.minimizingFiniteLabel_tendsto_zeroline 24Source documentation
The canonical choices inherit any vanishing competitor bound.
theorem minimizingFiniteLabel_tendsto_zero {L : Type*} [Fintype L] [Nonempty L] (cost : ℕ → L → ℝ) (hnonneg : ∀ n a, 0 ≤ cost n a) (a : ℕ → L) (ha : Tendsto (fun n => cost n (a n)) atTop (𝓝 0)) : Tendsto (fun n => cost n (minimizingFiniteLabel (cost n))) atTop (𝓝 0) := … - theorem
OAI.PolycyclicRecognition.linear_profile_distance_le_twice_affineline 33Source documentation
Subtracting the value at the identity costs at most a factor of two in the uniform norm. This removes arbitrary offsets from a comparison.
theorem linear_profile_distance_le_twice_affine {Z W : Type*} [TopologicalSpace Z] [CompactSpace Z] [NormedAddCommGroup W] (A B : C(Z, W)) (z₀ : Z) (hA : A z₀ = 0) (hB : B z₀ = 0) (b c : W) : dist A B ≤ 2 * dist (A + ContinuousMap.const Z b) (B + ContinuousMap.const Z c) := by … - theorem
OAI.PolycyclicRecognition.exists_finite_affine_label_separationline 56Source documentation
Finitely many distinct centered profiles have a uniform positive separation, even after arbitrary offsets are added.
theorem exists_finite_affine_label_separation {Z W L : Type*} [TopologicalSpace Z] [CompactSpace Z] [NormedAddCommGroup W] [Fintype L] [Nonempty L] (A : L → C(Z, W)) (hA : Function.Injective A) (z₀ : Z) (hzero : ∀ a, A a z₀ = 0) : ∃ ε : ℝ, 0 < ε ∧ ∀ a a' b b', dist (A a + ContinuousMap.const Z b) (A a' + ContinuousMap.const Z b') < ε → a = a' := by …